Secondary Restoration of Islanded AC Microgrids under Neural Inverse Optimal Control with Genetic Algorithm

A distributed neural inverse optimal controller is proposed as a secondary recovery for controlling frequency and voltage of inverter-based islanded microgrids. The proposed controller is further improved on the basis of the conventional droop control. The neural network trained by extended Kalman filtering (EKF) is used to approximate the unknown microgrid model. Next, the network model is used to implement the corresponding inverse optimal controller. Meanwhile, a genetic algorithm is used to optimize the parameters of the neural inverse optimal controller so as to improve the performance of the controller and achieve the overall optimization. The designed control system is in distributed form and only information from adjacent distributed generators is required. A four-unit system simulation is performed to verify the effectiveness of the controller.